The information required for adequate diagnosis, treatment and monitoring of cancers is so complex that a panel of measurements, used in sum, may provide the best answers. The concept is embodied in SELDI-TOF mass spectrometric (MS) peptide profiling, an emerging technique for serum based cancer detection. Even though SELDI has thus far only produced low complexity spectra, the patterns, when analyzed as groups, have the potential to create learning algorithms with diagnostic accuracies as good as or better than conventional biomarkers. We have developed a system to capture peptides on magnetic reversed-phase beads, followed by MALDI-TOF MS, to yield increasingly complex, yet very reproducible patterns. This has clear advantages, as more displayed peptides provide more opportunity to select unique patterns ('barcodes') for cancer subtypes and stages, and to predict and monitor clinical outcome. Extreme care has also been taken to standardize specimen collection, handling and storage to avoid the introduction of artifact. Pilot projects at MSKCC with a variety of malignancies suggest that peptide patterns thus obtained appear to hold information that may have direct clinical utility. The goals of this project are to (i) automate our prototype serum peptide profiling platform and implement machine learning methods that use the resulting peptide patterns ('barcodes') for sample classification [R21]; and (ii) to test the 'barcode diagnostic' model in a high-throughput setting, using well defined and carefully observed groups of thyroid carcinoma patients [R33]. R21 aim one is to automate serum sample processing and analysis; aim two is to automate all data processing, to examine pattern selection and sample class prediction methods, and to integrate all software platforms; aim three is to develop routine MALDI-TOF/TOF tandem MS sequencing of 'barcode' peptides. R33 aim one is to define reproducibility of serum patterns in patients with thyroid disease; aim two is to determine barcodes that can distinguish patients with thyroid cancer from those with benign thyroid nodules; aim three is to assess if serum peptidome barcodes can identify occult metastasis in a large group of thyroid cancer survivors. [unreadable] [unreadable]